IMO, they, like many other companies, were doomed by the constant chase for growth. Once they had a large share of "have a milk-coffee drink in a nice lounge" market, growth slowed. But having a large market share, good margins and growth that is the same as population/gdp (+/-) is just not acceptable.
So they try to find a way to get more growth, even if it changes and perhaps kills what the business was.
Around 2000 the founder stepped away, and MBAs brought in automated machines. They were more efficient and consistent at making the drinks than the baristas, and business tanked. The founder came back in 2008, got rid of the machines, and brought the baristas back. Business took off again.
In the context of AI automation I keep coming back to "cute Starbucks barista" as the archetypal automation-proof job. Because the job isn't producing the beverage, but the little moment of human interaction. (Especially these days, when not much of it remains!)
Same goes with supermarket checkout. I noticed many people intentionally take the line where the human scans your stuff. They enjoy it!
Unfortunately many zoomers do not appear to have been informed of this fact, and will give you a worse experience, "humanity wise", than the self-check out machine!
When you treat your job as robotic, aside from making the experience worse for all involved, you are also competing with actual robots, i.e. competing on speed, price and consistency, which is not a great place for a human to be.
I'm assuming you're talking about those Clover machines. They were really, really good and well designed IIRC. Trying to automate the barista with them; well, that's where they messed up!
So they try to find a way to get more growth, even if it changes and perhaps kills what the business was.